Using Discrete AI Features to Improve a Moving Average Crossover Strategy

Summary

This article compares a EURGBP daily moving-average crossover system with an AI-assisted version. The baseline combines exponential moving averages, a stochastic oscillator, and ATR-based exits. The proposed method trains models on earlier data to estimate how discrete indicator states relate to subsequent price changes, then uses those estimates to inform trading decisions. Training data ends before the stated backtest period, avoiding direct overlap between the two samples.

The author reports that the revised approach improved the listed performance measures in the backtest, including profitability and Sharpe ratio, and that it produced a majority of winning trades. The baseline performed poorly in the same test. These are reported results for one currency pair, timeframe, and historical interval; the excerpt does not provide enough detail to assess robustness across markets, costs, or alternative periods. The article also cautions that dummy encoding may not suit every market. Its main contribution is a feature-engineering framework, not proof that AI improves crossover strategies generally.

Key ideas

• The baseline system combines two exponential moving averages with a stochastic filter and ATR-based exits.
• The AI approach encodes indicator conditions as discrete states and learns their relationship to later price changes.
• Training and backtest periods are separated to reduce direct in-sample evaluation bias.
• The author reports improved backtest metrics for the AI-assisted system on EURGBP daily data.
• Results from one market and test interval do not establish general performance, and dummy encoding may not transfer to other settings.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.